Band structure fails to load if vasprun is gzipped and KPOINTS isn't

[ajjackson]

System

• Pymatgen version: 2018.7.23
• Python version: 3.6
• OS version: Linux Mint 17.3 64-bit

Summary

• Using the BSVasprun class to import a band structure calculation
• If vasprun.xml and KPOINTS files are present, it loads ok
• If both vasprun.xml and KPOINTS files have been gzipped (i.e. vasprun.xml.gz), it loads ok
• If only one file has been gzipped, an error is raised complaining about lack of KPOINTS file. This is quite disorientating if there is in fact a KPOINTS file...

Example code

gzip vasprun.xml

from pymatgen.io.vasp.outputs import BSVasprun
vr = BSVasprun('vasprun.xml.gz')
vr.get_band_structure(line_mode=True)

Error message

  File "/home/adam/.local/lib/python3.6/site-packages/pymatgen/io/vasp/outputs.py", line 749, in get_band_structure
    raise VaspParserError('KPOINTS needed to obtain band structure '
pymatgen.io.vasp.outputs.VaspParserError: KPOINTS needed to obtain band structure along symmetry lines.

Suggested solution

• This behaviour is caused by the following line in pymatgen.io.vasp.outputs:

        if not kpoints_filename:
            kpoints_filename = self.filename.replace('vasprun.xml', 'KPOINTS')

Presumably the "replace" expression is intended to maintain the first part of the path (i.e. the parent directory). However, if self.filename was /some/path/to/vasprun.xml.gz then the replace operation will leave /some/path/to/KPOINTS.gz.

• Suggested change

        if not kpoints_filename:
            vasprun_dir = os.path.abspath(os.path.join(self.filename,
                                                       os.path.pardir))
            for kpoints_filename in ('KPOINTS', 'KPOINTS.gz'):
                if os.path.exists(os.path.join(vasprun_dir, kpoints_filename)):
                    kpoints_filename = os.path.join(vasprun_dir,
                                                    kpoints_filename)
                    break
            else:
                kpoints_filename = ''

[computron]

Not sure why this is assigned to me (pretty sure I had no part in the original code)?

But anyway, I am not sure what os.path.pardir refers to in the above example.

I would suggest to just use monty zpath (https://pythonhosted.org/monty/_modules/monty/os/path.html)

Maybe something like this:

from monty.os.path import zpath

if not kpoints_filename:
    kpoints_filename = zpath(os.path.join(os.path.dirname(self.filename), "KPOINTS"))

It's a bit convoluted but should work (?)

[computron]

nevermind just understood the os.path.pardir as ".." - but anyway would probably suggest the code snippet above if it works

[mkhorton]

@ajjackson would you be happy to submit a PR to fix using monty's zpath ? (monty is already a dependency of pymatgen, zpath handles a range of compressed formats so would be a more general solution)

[ajjackson]

Can put that together, sure. This seems to be exactly what zpath is for.

[mkhorton]

Many thanks :)

[shyamd]

Fixed in in #1272